ep/l014106/1 supergen wind hub win… · 0 the wind speed, athe area of an anular ring of the...

EP/L014106/1 Supergen Wind Hub

Sustainable Power Generation and Supply

- Wind Energy Technologies

Deliverable 5.1.2

Report on assessment of Stress concentration method for

innovative blade design

Delivered by: 1

STFC Rutherford Appleton Laboratory

Author(s): Blanca Pascual1

Delivery date: 31st May 2017

Distribution list: Supergen Wind Consortium

Version No Status Date Checked by

1 Draft 31st May 2017 T.Davenne and

J.Halliday

Sponsored by

1

1 Introduction

The UK has a target of over 30% energy from renewables by 2020 to reduce its CO2 emis-sions. For this, onshore and offshore wind have to increase their capacity, from 10.1GW and5.3GW in may 2017, respectively to 13GW and 18GW by 2020. The SUPERGEN Wind En-ergy Technologies Consortium is a UK wind energy research consortium which was originallyestablished by the EPSRC on 23rd March 2006 as part of its Sustainable Power Generationand Supply(SUPERGEN) programme. Its aim is to develop academic, industrial and policylinkages and to lead the technology strategy for driving forward UK wind energy research andfor exploiting the research outcomes. This in turn should address the medium term challengesof scaling up to multiple wind farms, considering how to better build, operate and maintainmulti-GW arrays of wind turbines whilst providing a reliable source of electricity whose charac-teristics can be effectively integrated into a modern power system such as that in the UK. Theconsortium is led by Strathclyde University, but also includes Durham, Loughborough, Cran-field, Manchester, Oxford, Surrey, Bristol, Imperial, Dundee and STFC, with expertise in windturbine technology, aerodynamics, hydrodynamics, materials, electrical machinery, control, re-liability and condition monitoring, and has the support of 18 industrial partners, includingDNV GL and the Offshore Renewable Energy Catapult. Its activities were renewed in 2014,and is funded to 2019.

This report is related to work package 5.1, on innovative blade design for improved loadcontrol, where the use of sub-modelling is explored in ABAQUS to determine the extent towhich this might improve the modelling and representation of the structural loads withoutadding unduly to processing time.

This report studies the use of sub-modelling in ABAQUS to improve the representation ofstresses without adding unduly to the processing time.

The operating lifetime of wind turbine blades is controlled by the fatigue properties of thecomposite materials they’re made of. These materials are generally provided as fabrics (non-crimp fabrics, uniweave, woven), i.e. they have a 2D nature that is reconfigured into 3D shapeswhen the blade and its internal features (e.g. spars) are formed. Methods to assess fatigue lifedamage for 2D composites with in-plane loading have been the subject of several studies [3]but in regions such as the T-joint joining the blade spar to the aerodynamic profile, throughthickness effects are non-negligible and 3D stress states have to be taken into account.

The initiation of damage in these regions can occur well before what would be expected inthe base material subjected to in-plane loading. Numerical studies of these geometry-relatedfatigue issues are generally performed on test models of dimensions in the order of 0.1m [4, 5,7], using a very fine solid elements mesh. Wind turbine blade FE models available for academicpurposes have lengths two to three orders of magnitude larger than these T-joint test samples:e.g. the DTU 10MW turbine is 86.5m long [1]. These blade FE models use shell elements.Embedding the T-joint fine solid mesh within the shell mesh, so as to obtain an accurate stressassessment, could be both computationally challenging and need a detailed knowledge of thegeometry.

Here, the method proposed in the previous report [6] is further explored to use different testcases and vectors for the decomposition.

2 Validation models loading and geometry

2.1 Blade element momentum loads

In the previous report [6], a long extruded profile blade model was constructed using theideal aerodynamic section forces from blade element momentum (BEM), and these forces were

2

applied as pressures. Gravity and centrifugal forces were also applied. The section forces fromBEM are [2]

Fy = 2ρa(1 − a)v20A/3 Fx = 2ρ(1 − a)v0AΩa′R/3 a = 1/3 a′ =a(1 − a)

λR/Rmax

(1)

with ρ = 1.225kg/m3 the air density, v0 the wind speed, A the area of an anular ring ofthe actuator disc considered (A = π(R2

e − R2i )), a the the axial flow induction factor that

maximizes the power coefficient when the turbine is modelled as an actuator disc (e.g. [2]),a′ is the tangential flow induction factor, Ω is the angular velocity, R the radius of the ringconsidered, Rmax is the total radius of the rotor, λ = Ω ∗ Rmax/v0 is the tip speed ratio. Theblade is also subjected to centrifugal load (Ω = 2.8′rad/s) and gravity load, with the gravitypointing in the -z direction, the root of the blade located at Z=0, and the tip at Z=-25m.

2.2 Blade model geometries used for validation

Two simplified blade models are used to validate results, the extruded profile used for the testcases, and a blade using the same profile but following a given chord and twist distributionalong the length of the blade. This second blade was constructed because the geometry of thetest cases and of the long extruded profile blade used in the previous report [6] is the same,and the method should be tested with a different geometry to assess both that it can be usedwith a different geometry and that it is suitable for blades with changes in twist and chord.

2.2.1 Long Extruded profile

The geometry of this model is the one used to benchmark the results of the test cases in theprevious report [6]. In that report [6], the comparison results were not acceptable, as thestresses from the loaded blade and from the proposed method were very different. Here, it hasbeen decided to use the same geometry for the test cases described in subsection 3.1 and forthe first validation model. The loading in this validation model attempts to simulate a rotationof the blade. This load includes a gravity load that changes its angle, so as to simulate therotation, a centrifugal load, and a pressure distribution displayed in Figure 1b equivalent tothe BEM loading. The test cases and the validation model are studied in different steps ofthe same analysis, as they both use the same mesh, so that the test cases are computed insteps 1 to 12 and the rotation is studied in the last step. The model includes three kinematiccouplings, displayed in figure 2. For all the steps (test cases and rotation), the control node ofthe kinematic coupling at the root (see Figure 2a) has been fixed. For the test cases, differentpressures and displacements are applied in each test case, and these different displacements areapplied to the control nodes of the other kinematic couplings included in the model. Theseother kinematic couplings are located at the tip of the model (Figure 2b) and at an intermediatering of nodes (Figure 2c). In kinematic couplings, the nodes are slave to the translations androtations of the control node of the coupling. The elements where pressures are applied arehighlighted in Figures 2d and 2e.

3

(a) Long extruded profile mesh detail

(b) Long extruded profile pressure distribution

Figure 1: Long extruded profile geometry

(a) Root nodes (b) Tip nodes (c) Mid-ring nodes

(d) Elements for Fy pressure (e) Elements for Fx pressure

Figure 2: Nodes and elements where BCs/loads are applied to the extruded profile blade model

2.2.2 Profile following chord/twist distribution

This geometry uses the same profile, but it is scaled to modify the chord, and twisted along thelength of the blade. The wind energy handbook [2] provides a simplified blade geometry usingprofile NACA 4412, provided that the drag of the profile is ignored, ideal distributions of a anda′ are used, the lift coefficient maximizes lift/drag ratio Cl/Cd at µ = 0.7 and µ = 0.9 spanpoints, and assuming that the chord cu is linear with the length (straight line drawn throughthe 0.7 and 0.9 span points) of the blade. Then the chord, using a constant value of Cl = 0.7,is given by

cuRmax

=8

9λ0.8

(2 − µ

0.8

) 2π

ClλN(2)

4

with µ = R/Rmax, N the number of blades (here we assume three blades) and we assume theaerodynamic properties are as in NACA 4412, i.e. Cl = 0.7 for an angle of attack α = Cl/0.1−4(in degrees), that is constant along the blade length. The distribution of the lift coefficient canthen be obtained from the chord distribution

Cl =8

9

1

Ncuλ2π

√(2/3)2 + λ2µ2

[1 + 2

9λ2µ2

]2 α = Cl/0.1 − 4 (3)

and α is the angle of attack. The local inflow angle φ at each blade station varies along theblade is

tanφ =1 − a

λµ(1 + a′)=

1 − 1/3

λµ(

1 + 29λ2µ2

) β = φ− α (4)

So that the β is the twist of the blade. Here, the profile used is the one described in section2.2.1 instead of NACA 4412, but the chord and twist distribution provided in Equations 2 and4 are used. The pressures applied on the leading edge/trailing edge surfaces (PLE−TE) and topand bottom surfaces (PT−B), lead to the same aerodynamic forces as the ones obtained fromBEM in equation 1. The new pressures are now applied on surfaces that have been twistedby an angle β, and, for the leading and trailing edge surfaces, these surfaces have also beenscaled with the chord, and are equivalent to forces F ′y aligned with the LE-TE axis, and F ′x,perpendicular to this one. These forces can be related to Fy and Fx in equation 1 throughF ′x = cosβFx − sinβFy, F

′y = sinβFx + cosβFy

(a) Twist of the blade, seen from the root (b) Chord of the blade

Figure 3: Twist and chord of the blade

3 Method and test cases

The method was described in the previous report [6], and consists in the following steps

1. Create test cases using shell elements, all using the same geometry but different loads.The test cases used are further explained in subsection 3.1.

2. Extract vectors of data v for each of the test cases. Different vectors can be used withthis method, the ones perused here are described in subsection 3.2.

3. Create solids submodels for the T-joint region of the test cases, so as to have a detailedrepresentation of the stresses at the T-joint for the test cases.

4. Create a matrix whose columns are the data vectors from the test cases obtained in step2, i.e. V = [v1, . . . ,vi]. Calculate its Moore-Penrose pseudo-inverse V−1MP

5. From the blade shell model subjected to aerodynamic, centrifugal, gravity, gyroscopicloading: obtain a vector of data vg, similar to the one from the test cases in step 2.

5

6. Pre-multiply the vector from Step 5 by the Moore-Penrose pseudo-inverse of Step 4. Thisresults in a vector of coefficients α = V−1MPvg.

7. Combine the test cases submodels using the vector of coefficients α from Step 6. Thisresults in detailed stresses around the T-joint.

3.1 Long extruded profile test models

The geometry and couplings of this test model have been described in subsection 2.2.1. For thetest cases, data is extracted to form the vectors in Step 2 at the integration point of elementshighlighted in Figure 4. These elements are of size t× t (where t is the thickness of the shell),so that data at distances 0.5t, 1.5t and 2.5t from the T-joint can be extracted easily

Figure 4: Elements where stresses and strains are extracted

The test cases used with this geometry all have the control node of the kinematic couplinglinking nodes at the root (Figure2a) fixed in all its degrees of freedom. The other loading andBCs applied in each test case are

1. Gravity in Z direction: a distributed gravity load in the -Z direction is applied to allelements.

2. Gravity in X direction: a distributed gravity load in the X direction is applied to allelements.

3. Gravity in Y direction: a distributed gravity load in the Y direction is applied to allelements.

4. Centrifugal load: assumes the rotation speed is 2.8rad/s (tip speed ratio of 7, wind speedof 10m/s)

5. Pull-out: applies a pressure of 10Pa on the elements in Figure 2d

6. Pinch sides: Applies a pressure of 10Pa on the elements in Figure 2e

7. Tip z displacement: the control node of the kinematic coupling at the tip, controlingnodes in Figure 2b is displaced 0.1m in the Z direction.

6

8. Tip x displacement: the control node of the kinematic coupling at the tip is displaced0.1m in the X direction.

9. Tip y displacement: the control node of the kinematic coupling at the tip is displaced0.1m in the Y direction.

10. Tip Rz rotation: the control node of the kinematic coupling at the tip, is rotated 0.1radin the Y direction.

11. Mid ring x displacement: the control node of the kinematic coupling controlling nodesfrom Figure 2c is displaced 0.1m in the X direction.

12. Mid ring y displacement: the control node of the kinematic coupling controlling nodesfrom Figure 2c is displaced 0.1m in the Y direction.

The displacements of the above test functions are given in figure 5

7

(a) Gravity in Z (b) Gravity in X (c) Gravity in Y

(d) Centrifugal load (e) Pull-out (f) Pinch-sides

(g) Tip Z displacement (h) Tip X displacement (i) Tip Y displacement

(j) Tip Z rotation (k) Mid ring X displacement (l) Mid ring Y displacement

Figure 5: Extruded profile test cases displacement

3.2 Vectors of the subspace used in the proposed decompositionmethod

In the previous report [6], six types of vector were proposed as bases of six correspondingsubspaces. Some of these vectors are no longer reported, and some new vectors are described,namely Vector 3 and Vector 5. The results reported in section 4 will calculate the different α

8

vectors based on the bases obtained from

Vector 1 Corresponds to the first vector discussed in the previous report [6]. Vector obtainedfrom shell element membrane strains and curvatures in local 1 and 2 axis (outputs‘SE1’,‘SE2’, ‘SE3’, ‘SK1’, ‘SK2’, in abaqus) are extracted at 0.5t and 1.5t, i.e. for anelement on the outer skin and right side of the shear web, and at a distance 1.5t ofthe T-joint, a vector v′OSR,1.5 = [SE1, SE2, SE3, SK1, SK2] is created. This leadsto 5 (strains/curvatures) x 2 (distances from T-joint) x3 (sides around the T-joint:outer skin right, outer skin left and shear web) =30 datum to be used in each datavector of the test cases/global model, so that the data vector is given by

v =

vOSR,0.5vOSR,1.5vOSL,0.5vOSL,1.5vSW,0.5vSW,1.5

with vOSR,0.5 =

SE1OSR,0.5SE2OSR,0.5SE3OSR,0.5SK1OSR,0.5SK2OSR,0.5

(5)

It is noted that results obtained with this basis are only displayed for the straightblade at location 1 in Figure 8, where this method performs poorly in comparisonwith the other ones.

Vector 2 Corresponds to the fourth vector discussed in the previous report [6]. Vector obtainedfrom shell membrane strains at the top and bottom of the shells (‘E11’, ‘E22’ and‘E12’ outputs in abaqus for the topmost and bottommost section points of the shells)at distances 0.5t and 1.5t of the T-joint. This leads to 3 (strains) x2 (top/bottomshell location) x 2 (distances from T-joint) x3 (shells around the T-joint) =36 datumto be used in each data vector of the test cases/global model.

v =

vOSR,topvOSR,bottomvOSL,topvOSR,bottomvSW,topvSW,bottom

with vOSR,top =

E11OSR,top,0.5tE22OSR,top,0.5tE12OSR,top,0.5tE11OSR,top,1.5tE22OSR,top,1.5tE12OSR,top,1.5t

(6)

Vector 3 New vector introduced in this report. Vector obtained from shell membrane strainsat the top and bottom of the shells at distances 1.5t and 2.5t from the T-joint. Thedata vector has a size of 3x2x2x3=36

v =


with vOSR,top =

E11OSR,top,1.5tE22OSR,top,1.5tE12OSR,top,1.5tE11OSR,top,2.5tE22OSR,top,2.5tE12OSR,top,2.5t

(7)

Vector 4 Corresponds to the sixth vector discussed in the previous report [6]. Vector obtainedfrom shell membrane strains at the top and bottom of the shells at distances 0.5t,

9

1.5t and 2.5t from the T-joint. The data vector has a size of 3x2x3x3=54

v =


with vOSR,top =

E11OSR,top,0.5tE22OSR,top,0.5tE12OSR,top,0.5tE11OSR,top,1.5tE22OSR,top,1.5tE12OSR,top,1.5tE11OSR,top,2.5tE22OSR,top,2.5tE12OSR,top,2.5t

(8)

Vector 5 New vector introduced in this report. This vector tries to establish if adding stressesto the vector of data improves the accuracy of the method. Vector obtained fromshell membrane strains and shell stresses at the top and bottom of the shells (‘E11’,‘E22’, ‘E12’, ‘S11’, ‘S22’ and ‘S12’ outputs in abaqus for the topmost and bottommostsection points of the shells) at distances 0.5t and 1.5t of the T-joint. This leads to6 (strains and stresses) x2 (top/bottom shell location) x 2 (distances from T-joint)x3 (shells around the T-joint) =72 datum to be used in each data vector of the testcases/global model.

v =


with vOSR,top =

E11OSR,top,0.5tE22OSR,top,0.5tE12OSR,top,0.5tE11OSR,top,1.5tE22OSR,top,1.5tE12OSR,top,1.5tS11OSR,top,0.5tS22OSR,top,0.5tS12OSR,top,0.5tS11OSR,top,1.5tS22OSR,top,1.5tS12OSR,top,1.5t

(9)

3.3 Submodels

The T-joint geometry used here is the same as the one used in [7, 5], and its geometry andlayup are described in Figure 6. The orientation is such that the second axis corresponds tothe direction of the fibre. The material properties of the uniaxial composite are: elastic moduliE1 = E3 = 9 GPa, E2 = 38 GPa; Poisson’s ratios ν13 = ν23 = 0.3, ν12 = 0.071; and the shearmoduli G12 = G23 = 3.6 GPa, G13 = 3.46 GPa. Due to the elements orientation, stresses S11are perpendicular to the fibre, S22 are in the direction of the fibre and S12 are shear stresses.

10

Figure 6: T-joint geometry

The submodels for the test case and the long extruded profile blade have the same geometry.For the blade with twist/chord variation, the T-joint submodel nodes had a change of coordi-nates, so that for each Z coordinate, all the nodes had their x and y coordinates translated tofollow the twist description. This maintains the thickness of the layers, the local radius of theT-joint, the mesh size, and allows to follow the geometry of the blade. Figure 7 provides themesh for the submodel of the extruded profile blade, and the elements where stresses have beenextracted for each ply so as to obtain results in section 4. For all the plies in the shear web,the distance across the radius of the bend is scaled so that they have the same total distance,and the distances in Figure 7 are the ones used as X coordinate in the figures of section 4.

(a) Submodel elements wheredata is extracted

(b) Distances along the outerskin elements

(c) Distances along the shearweb elements

Figure 7: Submodel elements where stresses are extracted, and distances used to plot the results

4 Results

Initially, to check that the method works, a model using the extruded profile geometry wascreated, where the first step applies gravity in the -Z direction, the second step applies gravityin the X direction, and the third step applies -11 times the gravity in Z direction and -4 timesthe gravity in the X direction.The coefficients α1 = −11, α2 = −4 were retrieved accuratelyusing the method with the first two steps used as the only two test cases.

Four sets of results are given. Firstly, submodel results are compared for the model withsame geometry as the test cases at the same location where the data is extracted (at Z=-6.18m),displayed in black in Figure 8. Then, results are compared at a different Z coordinate within

11

the T-joint (Z=-12.38m), displayed in red in Figure 8. Then, at the first Z coordinate on thetwisted blade from subsection 2.2.2, and finally at the second Z coordinate on the twisted blade.

Figure 8: Locations of the submodels in the straight blade (Location 1 in black, location 2 inred)

4.1 Results from extruded profile blade at same location

The extruded profile is used for both the test cases and the validation of the results, and datais extracted in both cases at Z=-6.18m. In the last step of the analysis, the blade starts theanalysis at the bottom of the rotation, and the rotation is divided into 20 frames, where theanalysis is static and the only change between the different frames is the angle of the gravityvector. Results from the 7th frame of the analysis are used for the comparison of stressesobtained directly from the rotating blade submodel and from the method proposed in section3.

Some examples of the comparison between the stresses on the submodels are given in figures9, 10, 11, 12, 13 and 14. The red line in each figure provides the results of the submodel of theblade model subjected to distributing loading (BEM, gravity, centrifugal). The remaining linescombine the results of the submodels from the test cases, and differ in the basis of the subspaceused to obtain the α coefficients. The cyan, orange, navy and purple lines use, respectively, theshell strains at distances 0.5t and 1.5t ( Vector 1 ); the strains at top and bottom of the shellsat distances 0.5t and 1.5t ( Vector 2); the strains at top and bottom of the shells at distances0.5t, 1.5t and 2.5t ( Vector 4 ); the strains at top and bottom of the shells at distances 1.5t and2.5t ( Vector 3 ); The turquoise line uses strains and stresses at top and bottom of the shellsat distances 0.5t and 1.5t ( Vector 5 )

12

(a) Bottom of ply-1

(b) Middle of ply-5

(c) Middle of ply-6

(d) Top of ply 15

Figure 9: S11 submodel stresses at different plies of the outer skin. The ply is indicated in thecaption of each subfigure. Stresses are in Pa, distances in m.

13

(a) Bottom of ply-1

(b) Top of ply 5

(c) Top of ply 10

(d) legend

Figure 10: S11 submodel stresses at different plies of the shear web. The ply is indicated inthe caption of each subfigure. Stresses are in Pa, distances in m.

14

(a) Bottom of ply 1

(b) Middle of ply 5

(c) Middle of ply 6

(d) Top of ply 15

Figure 11: S22 submodel stresses at different plies of the outer skin. The ply is indicated inthe caption of each subfigure. Stresses are in Pa, distances in m.

15

(a) Bottom of ply 1

(b) Top of ply 5

(c) Top of ply 10

(d) legend


16

(a) Bottom of ply-1

(b) Middle of ply-5

(c) Middle of ply-6

(d) Top of ply-15


17

(a) Bottom of ply-1

(b) Top of ply-5

(c) Top of ply-10

(d) legend


Results for the through thickness stresses (S33) are provided below

18

(a) S33 submodel stresses next to the glue on the outer skin side of the glue

(b) S33 submodel stresses next to the glue on the shear web side of the glue

Figure 15: Submodel through thickness stresses. Stresses are in Pa, distances in m.

4.2 Results from extruded profile blade at location 2

Results for the location 2 in figure 8 are displayed below. The line providing results basedon the shell strains (Vector 1 in subsection 3.2), has been removed as the other vectors usedto obtain α coefficients consistently performed better. Results remain good at the secondlocation, particularly at the region where the stresses have been extracted to obtain the vectorsin subsection 3.2 (region closest to the T-joint).

19

(a) Bottom of ply-1

(b) Middle of ply-5

(c) Middle of ply-6

(d) Top of ply 15


20

(a) Bottom of ply-1

(b) Top of ply 5

(c) Top of ply 10

(d) legend


21

(a) Bottom of ply 1

(b) Middle of ply 5

(c) Middle of ply 6

(d) Top of ply 15


22

(a) Bottom of ply 1

(b) Top of ply 5

(c) Top of ply 10

(d) legend


23

(a) Bottom of ply-1

(b) Middle of ply-5

(c) Middle of ply-6

(d) Top of ply-15


24

(a) Bottom of ply-1

(b) Top of ply-5

(c) Top of ply-10

(d) legend



25


(b) S33 submodel stresses next to the glue on the shear web side of the glue

Figure 22: Submodel through thickness stresses. Stresses are in Pa, distances in m.

4.3 Results from twisted blade blade at location 1

Results for the first location studied on the twisted blade are displayed below. The line providingresults based on the shell strains (Vector 1 in subsection 3.2), has been removed as the othervectors used to obtain α coefficients consistently performed better. Results remain good atthe first location of the twisted blade, particularly at the region where the stresses have beenextracted to obtain the vectors in subsection 3.2 (region closest to the T-joint).

26

(a) Bottom of ply-1

(b) Middle of ply-5

(c) Middle of ply-6

(d) Top of ply 15


27

(a) Bottom of ply-1

(b) Top of ply 5

(c) Top of ply 10

(d) legend


28

(a) Bottom of ply 1

(b) Middle of ply 5

(c) Middle of ply 6

(d) Top of ply 15


29

(a) Bottom of ply 1

(b) Top of ply 5

(c) Top of ply 10

(d) legend


30

(a) Bottom of ply-1

(b) Middle of ply-5

(c) Middle of ply-6

(d) Top of ply-15


31

(a) Bottom of ply-1

(b) Top of ply-5

(c) Top of ply-10

(d) legend



32


(b) S33 submodel stresses next to the glue on the shear web side of the glue.Stresses are in Pa, distances in m.

Figure 29: Submodel through thickness stresses

4.4 Results from twisted blade blade at location 2

Results for the second location studied on the twisted blade are displayed below. The lineproviding results based on the shell strains (Vector 1 in subsection 3.2), has been removed asthe other vectors used to obtain α coefficients consistently performed better. Results remaingood at the second location, particularly at the region where the stresses have been extractedto obtain the vectors in subsection 3.2 (region closest to the T-joint).

33

(a) Bottom of ply-1

(b) Middle of ply-5

(c) Middle of ply-6

(d) Top of ply 15


34

(a) Bottom of ply-1

(b) Top of ply 5

(c) Top of ply 10

(d) legend


35

(a) Bottom of ply 1

(b) Middle of ply 5

(c) Middle of ply 6

(d) Top of ply 15


36

(a) Bottom of ply 1

(b) Top of ply 5

(c) Top of ply 10

(d) legend


37

(a) Bottom of ply-1

(b) Middle of ply-5

(c) Middle of ply-6

(d) Top of ply-15


38

(a) Bottom of ply-1

(b) Top of ply-5

(c) Top of ply-10

(d) legend



39


(b) S33 submodel stresses next to the glue on the shear web side of the glue.Stresses are in Pa, distances in m.

Figure 36: Through thickness stress in the submodel

5 Conclusions

The proposed method seems to lead to good results in the vicinity of the T-joint where thedata to compare shell results has been extracted. The agreement for the through thicknessstress is particularly good. These stresses are very relevant in the failure mechanism of T-jointsunder pull-out test loading [7]. The next steps to develop further the method could considerincluding the effect of the angle between the shear web and the outer skin, as this angle isnot always 90 degrees in a real blade. This could be performed by having the same test caseswith several angles, and then automatically identifying the angle in the blade from the mesh.The other topic that is of interest is the stiffness reduction of the joint as it gets damaged.A connector could be included in the blade shell model, joining the outer skin and the shearweb, that would model the damage. Research could be carried out on how to determine theproperties of the connectors to be used, so as to include a measure of the damage, and the rightstiffness properties of the T-joint with damage. The damage in the T-joint to be modelled by aconnector could encompass both the stiffness reduction with fatigue life in the composite itself,and also the fracture next to the glue reported in [7].

References

[1] C. Bak et al. “Design and performance of a 10MW wind turbine”. In: Danish Wind PowerResearch (2013).

40

[2] T. Burton et al. Wind Energy Handbook. Wiley, 2001.

[3] J. Degrieck and W. Van Paepegem. “Fatigue damage modelling of fibre-reinforced com-posite materials: review”. In: Applied Mechanics Reviews 54.4 (2001), pp. 279–300.

[4] H. Gulasik and D. Coker. “Delamination-debond behaviour of composite T-joints in windturbine blades”. In: The science of making torque from wind 2014, Journal of PhysicsConference Series 524.012043 (2014).

[5] Amirhossein Hajdaei. “Extending the fatigue life of a T-joint in a composite wind turbineblade”. PhD thesis. Manchester University, 2013.

[6] B. Pascual. EP/L014106/1, Deliverable 5.1.1: Report on Stress concentration estimationmethod for innovative blade design. Tech. rep. Supergen Wind Hub- Sustainable PowerGeneration and Supply-Wind Energy Technologies, 2017.

[7] Y.Wang et al. “Finite element analysis of composite T-joints used in wind turbine blades”.In: Plastics, Rubber and Composites 44.3 (2016), pp. 87–97.

41

ep/l014106/1 supergen wind hub win… · 0 the wind speed, athe area of an anular ring of the...

Documents